4. LS-DYNA vs. PAM-CRASH for whiplash simulation

Whiplash simulation is a niche within crashworthiness analysis where the two dominant explicit solvers—LS-DYNA and PAM-CRASH—take noticeably different architectural approaches. Both are explicit finite element codes with Lagrangian formulations, both support the BioRID-II and THUMS human body models, and both can reproduce the rear-impact pulse defined in Euro NCAP or IIHS procedures. Yet the way each solver handles contact, time integration defaults, and keyword semantics means that a model built for one cannot be dropped into the other without deliberate translation. This section examines where the two solvers diverge, how their keyword systems map onto each other, and what a practical conversion workflow looks like.

Where the Solvers Actually Differ

At the formulation level, LS-DYNA and PAM-CRASH share the same explicit central-difference time integration scheme, so the stability limit is governed by the same Courant condition: the critical time step is proportional to the smallest element characteristic length divided by the dilatational wave speed. In practice, however, the two codes apply different default mass-scaling and time-step controls. LS-DYNA uses *CONTROL_TIMESTEP with a default time-step scale factor of 0.9 and optional selective mass scaling via DT2MS. PAM-CRASH exposes equivalent control through the /TIME card family, where the time-step scale factor and mass-scaling options are set separately. A model that runs stably in LS-DYNA with a 0.9 scale factor may become unstable in PAM-CRASH if the corresponding card is left at its default, because the default scale factor and the treatment of rigid-body mass differ.

Contact treatment is the second major divergence. LS-DYNA's penalty-based contact (*CONTACT_AUTOMATIC_SURFACE_TO_SURFACE) and its tied contact (*CONTACT_TIED_SURFACE_TO_SURFACE) are widely used for the seat-back-to-torso and headrest-to-head interfaces in whiplash models. PAM-CRASH uses a comparable penalty formulation but organizes it under /CONTACT with type codes such as TYPE 34 (surface-to-surface) and TYPE 2 (tied). The friction model, the search algorithm for slave nodes, and the handling of initial penetrations are not identical. In whiplash models, where the headrest foam compresses against the head and the seat back rotates against the torso, small differences in contact stiffness can shift the peak head rotation angle by several degrees—enough to change a Euro NCAP whiplash rating.

Material models form the third difference. LS-DYNA's *MAT_LAW family (for example *MAT_024 for piecewise linear plasticity, *MAT_057 for low-density foam) has direct PAM-CRASH counterparts in the /MAT card set, but the parameter ordering and the definition of strain-rate curves differ. Foam materials are particularly sensitive: LS-DYNA's *MAT_057 expects a nominal stress–strain curve, while the PAM-CRASH equivalent may expect a true stress–strain curve or a different unloading flag. Copying numbers without checking the curve definition is a common source of error.

Keyword Mapping in Practice

The table below summarizes the most frequently encountered mappings when converting a whiplash deck. It is not exhaustive, but it covers the cards that most often cause a conversion to fail or produce divergent results.

A Workable Conversion Workflow

Converting a whiplash model between the two solvers is not a one-click operation. A reliable workflow proceeds in stages, with verification at each step.

1. Inventory the source deck. List every keyword card, material, contact definition, and boundary condition. Identify which cards have no direct counterpart—these are the ones that will require manual re-authoring.
2. Translate geometry and mesh. Both solvers accept Nastran or IGA format meshes, so the mesh itself usually transfers cleanly. Check for duplicate nodes, free edges, and element quality metrics after import, because the two codes have different default tolerances for element distortion.
3. Map materials and verify curves. For each material, confirm whether the stress–strain input is nominal or true, and whether the unloading behavior is defined explicitly or inferred. Foam and seat fabric materials are the most sensitive to this step.
4. Re-author contacts. Do not assume that a contact definition transfers. Rebuild each contact pair in the target solver, set the friction coefficients explicitly, and check the initial gap. Run a short settling simulation to confirm that no initial penetration exists.
5. Re-apply the pulse. The rear-impact pulse (typically a sled acceleration curve) must be redefined in the target solver's function format and applied to the correct node set or rigid body. Verify the pulse by plotting the prescribed velocity against time.
6. Run a short verification case. Before running the full whiplash simulation, run a 20–30 ms simulation with the head restrained and compare the seat-back rotation and headrest contact force against the source result. A discrepancy of more than 5–10% in these quantities indicates a mapping error.
7. Compare key whiplash metrics. Once the full simulation runs, extract the NIC (Neck Injury Criterion), head rotation angle, and head–headrest contact force. These are the quantities that determine the whiplash rating, and they are the most sensitive to solver differences.

Practical Example: Converting a BioRID-II Rear-Impact Deck

Consider a BioRID-II model with a seat back, headrest, and a prescribed sled pulse of 15 g peak acceleration over 80 ms. The source deck is in LS-DYNA with *MAT_057 for the headrest foam and *CONTACT_AUTOMATIC_SURFACE_TO_SURFACE for the head–headrest interface. After conversion to PAM-CRASH, the first run shows a 12% higher peak head rotation angle. The cause is traced to the foam material: the LS-DYNA deck used a nominal stress–strain curve, while the PAM-CRASH /MAT FOAM card expects true stress–strain. Converting the curve using the incompressibility assumption (true stress = nominal stress × (1 − nominal strain)) reduces the discrepancy to under 3%. This is a representative example of the kind of adjustment that conversion requires; actual results will depend on the specific model and pulse.

The two solvers are capable of producing equivalent whiplash results, but equivalence is achieved through deliberate mapping and verification, not through automatic translation. The engineer who understands where the defaults differ—time-step scaling, contact penalty, and material curve definitions—will convert models faster and with fewer surprises.